Un- reveals antonymy in the lexicon
Andrew Paczkowski
1 Introduction
Introductory textbooks explain the semantics of the un- prefix thus: when applied to an
adjective it has a negative meaning, and when applied to a verb it has a reversative meaning.
Thus un- creates antonyms from both complementary and gradable adjectives, as in true/untrue
and happy/unhappy. For verbs, the addition of un- is only able to create the reversative
antonyms as in lock/unlock, but never the conversative as in lend/borrow as opposed to
lend/*unlend, likely because antonyms of verbs like lend require an often unpredictable shuffling
of theta roles. Although this paper does not address verbs, I do hope that it will assist in that
investigation.
I hope to show that un- does not create antonyms by adding the notion of negation to a
positive form, but rather by adding the notion of antonymy itself. This description of the
behavior of un- accounts for certain gaps which on the surface appear to be the result of some
blocking effect.
Before beginning the analysis, I must point out that this paper makes no attempt to
explain various occurrences of the nonproductive version of un-, such as occurs in unrest,
unemployment, unbeliever, the recent back-formation unbelieve, press-related jargon words like
unhave, etc. These are affective innovations and historical residues of older usages, and are
therefore either no longer productive or only marginally productive.1
2 Un- the Non-Negator
1 For an interesting discussion on more marginal uses of un- in popular culture, see Ben Zimmer's article “The
Age of Undoing” in New York Times Magazine, 15 September 2009.
1
As noted above, un- operates without regard to gradability. A clearer pattern falls out
when one groups adjectives as either subsective, intersective, or privative. The three types can
be clarified quite simply by outlining what truth values each entails about the noun it describes.
(1)
subsective
This is a big beetle
→
This is a beetle.
↛
This beetle is big.
intersective
This is a kind man.
→
This is a man.
→
This man is kind.
privative
This is a fake teapot. ↛
→
This is a teapot.
This teapot is fake.
The crossed arrows indicate that the following statement is not entailed by the preceding
statement. Beetles, for instance, are small things compared to humans, but one may refer to a
beetle which towers over other beetles as a big beetle. It may be obvious—but I shall address it
here briefly—that there need only be three such categories, because an adjective which entailed
neither such truths would have no linguistic value.2
The grammaticality of privative adjectives prefixed with un- is remarkably uniform. We
might expect that it will create a form synonymous with that adjective's antonym, but this is not
the case.
(2)
fake
*unfake
2 If this is a blargle teapot entailed neither this is a teapot nor this teapot is blargle, what semantic value could
blargle have?
2
counterfeit
*uncounterfeit
fictitious
*unfictitious
imaginary
*unimaginary
In all these instances one might conclude that the prefixed form is blocked by a salient antonym
like real. The same explanation presents itself for subsective adjectives as well, because they are
all3 likewise ungrammatical when prefixed with un-.
(3)
big
*unbig
small
*unsmall
far
*unfar
near
*unnear
good
*ungood
bad
*unbad
tall
*untall
short
*unshort
In fact the only adjectives with which un- can be productive are intersective, though it does not
create grammatical forms with all intersective adjectives.
(4)
true
untrue
happy
unhappy
kind
unkind
3 Confusion may arise from certain subsective adjectives that have intersective homophones. The word intelligent,
for instance, can be subsective in the sense that this is an intelligent cow does not entail this cow is intelligent,
because here the word has a meaning similar to 'clever' or 'smart' (note: *unclever, *unsmart). But intelligent
can be intersective in the sense that this is an intelligent organism does entail this organism is intelligent,
because here the word means something like 'having a brain' or 'conscious.' Thus the formation unintelligent was
likely made from the intersective version of intelligent, and later extended as a synonym of stupid. I predict that
any other apparent counterexamples (except the affective formations mentioned in section I) will also exhibit
homophony with an intersective adjective.
3
clean
unclean
healthy
unhealthy
green
*ungreen
glassy
*unglassy
round
*unround
flat
*unflat
A basic look at the semantics of their components makes the ungrammaticality of forms like
*ungreen and *unround apparent. Here we must draw a clear distinction: it is from such forms
as these that I conclude that un- is not a negator except insofar as antonyms overlap with
negatives, and that the function of un- is to form antonyms rather than negations. A complex
idea like flat, for instance, can be negated, but does not have a contrary or gradable antonym.
What would be the antonym of flat? Is it round, lumpy, or perhaps pointy? There is no singular
opposition with which to draw distinction. In this way un- contrasts with non-, which does
indeed form negatives like non-flat, the meaning of which is allowed by the semantic value of
non- to encompass round and lumpy as well as pointy while committing to no particular one. In
a sense all these options are antonyms of flat, but the relationship is asymmetric, which is why
many restrict the term 'antonym' to those oppositions that are gradable and contrary.4 So perhaps
we should take it as a matter of course that the function of un- is to create not just an opposition,
but an antonym par excellence.
3 More Gaps: Markedness to the Rescue?
Many adjectives are not so easily explained. These are adjectives which fit the criteria
for production with un-: they are intersective and have singular antonyms, yet they do not create
4 Murphey 2003, 189; Lehrer and Lehrer 1982
4
grammatical forms when prefixed thus.
(5)
false
*unfalse
sad
*unsad
mean
*unmean
dirty
*undirty
sick
*unsick
Nor can we invoke the simple explanation that they are blocked by more salient antonyms,
because then we cannot account for grammatical forms such as in (4). Moreover, the failure of
that explanation in these instances undermines its ability to explain the ungrammatical forms in
(2) and (3).
Much has been said about the role of markedness in antonymy, but this will be incapable
of describing the behavior of un- with privative and subsective adjectives because in each of
these groups there are marked and unmarked forms, all equally shunned by un-. Nevertheless it
warrants investigation since it may account for the varying applicability of un- to intersectives.
But first we must explore the idea of markedness.
Since markedness is a linguistic term, its definition depends upon one's field (phonology,
syntax, etc.) and theory. Semantic markedness—this paper's concern—has been treated
according to its alignment with morphological complexity, syntactic distribution, the possibility
of further semantic differentiation, and prototypical versus nonprototypical concepts.5
The identification of semantic markedness with the expression of more or less of a quality goes
back to the original work on markedness done by Roman Jakobson.6 Lehrer captures the
5 Battistella 1990, 34-45
6 Battistella 1990, 28-33.
5
situation in the following quote:
“In looking at the evaluative or connotative 'meaning', the unmarked
member has a positive connotation and the marked member has a negative one.
Thus happy, clean and friendly, which come out as unmarked according to the
criteria listed above, have favourable connotations, while their antonyms, sad,
dirty and unfriendly have negative connotations.
Unmarked members of an antonym pair denote more of a quality, while
the marked member denotes less. According to the previous criteria, big, tall,
heavy and old are unmarked and indeed these terms denote more size, height,
weight and age than their corresponding antonyms small, short, light and young.”
(Lehrer 1985)
Reading over the words in (5), the question that appears most glaring is this: why should
negative evaluation correlate to markedness? It is, one must admit, an empirically tenuous
connection between, say, dirty and false. Lehrer's statement, “unmarked members of an antonym
pair denote more of a quality...” makes less sense when one considers more closely the examples
happy, clean, and friendly. Is cleanliness something that one has more of, and dirtiness what
encroaches in its absence? Perhaps in a very abstract sense, though the reasoning seems a bit
contrived. All the above approaches have their downfalls,7 and each would define markedness in
an unhelpfully different way. Therefore let it suffice to identify markedness with an intrinsic
affinity for asymmetry in usage. This intrinsicness, however, compels us to analyze markedness
as perhaps a lexical feature. But Murphy notes that an approach that would treat markedness as a
7 I highly recommend reading the criticism offered by Battistella 1990, on pp. 34-45.
6
lexical feature cannot apply to any contextually dependent markedness.8 One asks how good is
that book? but not #how good is that oil-spill? or #how good is your flu?. This approach also
fails to account for antonymy in general, since there are antonymic pairs like hot and cold (as
opposed to warm and cool) in which neither member is inherently marked with respect to the
other.9
Although markedness may provide clues in this investigation, I do not think it can be
made at all diagnostic. A word like problematic, for instance, is patently negatively evaluated,
and one would probably label it marked with regard to an antonym like easy. And despite the
word's morphological complexity, one may find occasion to produce the antonym unproblematic
in order to fill a perceived gap.10 Now problematic does denote more of a quality, and so may be
labeled unmarked according to that criterion, but how are we to determine when a form will
follow which criteria? I think that the behavior of un- can inform this discussion more than it
can be illuminated by it.
4 Antonymy in the Lexicon
The way I propose to analyze negative evaluation adjectives is to break down their
semantic values in terms of their positive evaluation counterparts. I believe the existence of a
morpheme whose most basic semantic value consists in specifying antonymy, namely un-,
supports the possibility that {ANTONYM} (hereafter abbreviated {ANT}) may exist
8 Murphy 2003, 186-188
9 Murphy 2003, 187
10 This brings up another point, namely that un- can be used to coerce gradable contrariety. The word pointy, for
instance, has no distinct antonym when applied to rocks (round? flat? dull?) but it could when applied to spears.
And in case one hesitates to use dull and sharp for something that is not quite a blade, ?unpointy could be
formed with a clear meaning in its context, where it would be opposed to pointed, 'having a point.' Similarly one
cannot describe a food as ?unsweet, because this does not indicate any particular opposition (salty? bitter?
sour?). But in the southern United States, unsweet tea denotes tea without sugar added. In the context of tea,
sweet is synonymous with sweetened. And so the addition of un- to this particular version of sweet by analogy
produces a synonym of unsweetened.
7
independently as a semantic value available for inclusion anywhere in the lexicon. Thus the
word true would have the semantic value {TRUE}, but false would have the value {ANT,
TRUE}, specifying it in the mental lexicon as the antonym of {TRUE}. So too would untrue
have the value {ANT, TRUE}. And thus the reason for uniformly preferring true to its would-be
synonym *unfalse, is that the former would have the simple value of {TRUE} while the latter
would heap semantic values together redundantly with the value of {ANT, ANT, TRUE}, which
is interpretable but inefficient. The answer then, if this analysis be correct, is a blocking effect
which compels speakers to avoid such inefficiency at the level of semantic composition. Now let
us turn our attention to privative and subsective adjectives.
This analysis accounts for privative adjectives easily. All of them can be identified
semantically as the asymmetrically abundant antonyms of real, since they draw their
characteristic shared feature through opposition to that concept. This gives us two reasons why
privatives avoid un-. First, they run into the same redundancy problem as the *unfalse example.
Second, they have the same asymmetry as flat, in that they do not have clear antonyms par
excellence. Therefore the privative category would seem, below the surface, to be derivative of
the intersective category. Alternatively, Partee considers privatives to be really subsectives
opposing such concepts as real.11 With that view, privatives can be dealt with implicitly in the
following excursus on subsective adjectives. In any case, privatives will be seen ultimately to
belong to one of the other two categories.
Subsective adjectives pose an interesting dilemma. Here we have a set of positive and
negative evaluation words, clear gradable and contrary antonyms par excellence, and yet no
possibility of affixation with un-. The reason is that subsective pairs locked in a catch-22 that
11 Partee 2010
8
admits of no such derivation. They are logically codependent because their opposition is what
gives them meaning; but they are semantically independent because deriving one from the other
would require the other to be logically independent. A subsective without an equal but opposite
subsective antonym cannot be a subsective at all. For instead of creating dichotomies
recursively, an unopposed subsective would create only a scale of degree, making it
indistinguishable from and therefore equal to an intersective. In order to use un- to create an
antonym to a subsective, it would have to be attached to one of an already existing pair of
subsectives, which would result in the same sort of indirectness of express that we encounter
with redundancy. For example, a pair like big and small may decompose into simply {BIG} and
{SMALL}. Thus *unsmall {ANT, SMALL}, meaning 'big,' and *unbig {ANT, BIG}, meaning
'small,' are not avoided merely because they are more complex than their alternatives big {BIG}
and small {SMALL}, but also because they circumvent another semantic prime. Compare the
following competing forms.
(6a)
*unfalse
{ANT, ANT, TRUE} competes with
true
{TRUE}
(6b)
untrue:
{ANT, TRUE}
competes with
false:
{ANT, TRUE}
(6c)
*unbig:
{ANT, BIG}
competes with
small:
{SMALL}
This reasoning accounts for all the data and, if it is correct, should help inform further
investigations of markedness.
Support for my breakdown may be found in the work of Wierzbicka and Goddard, who
include in their list of semantic primes the notions BIG and SMALL, as well as GOOD and
BAD, FAR and NEAR.12 But although they also include TRUE,13 one notices the absence of
12 Goddard and Wierzbicka 1994, 37-47
13 Goddard 2002
9
*FALSE. Their reasoning derives for the most part from cross-linguistic universality, which, if
it is correct, would have interesting implications for my analysis. Furthermore if my analysis is
correct, it may be necessary to add ANTONYM to their canon of primes. In any case, the next
step in investigating this phenomenon is to attempt to model similar patterns in languages
besides English.
5 Conclusion
The prefix un- can be applied to any adjective which is logically able to stand in a
relationship of antonymy par excellence with another adjective. The difficulty of using
markedness alone to explain the distribution of un- disappears when one considers its semantic
value as that of {ANTONYM} and that this same semantic value may be inherent in certain
intersective adjectives, which are thus stored in the lexicon as antonyms of other concepts14.
Antonymy in the mental lexicon thus accounts for the distribution of un-.
14 As an interesting mental exercise, consider the processing time required to draw out the meaning from phrases
like John lacks a dearth of paucity of books.
10
Works Cited
Battistella, Edwin L. 1990. Markedness: The Evaluative Superstructure of Language. New York:
State University of New York Press.
Goddard, Cliff. 2002. The search for the shared semantic core of all languages. Meaning and
Universal Grammar - Theory and Empirical Findings: Volume I, eds. Cliff Goddard and
Anna Wierzbicka. Amsterdam: John Benjamins. pp. 5-40.
Goddard, Cliff and Anna Wierzbicka. 1994. Introducing Lexical Primitives. Semantic and
Lexical Universals, eds. Cliff Goddard and Anna Wierzbicka. Philadelphia: John
Benjamins. pp. 31-54.
Lehrer, Adrienne. Sep., 1985. Markedness and Antonymy. Journal of Linguistics, 31(3):397-429.
Lehrer, Adrienne and K. Lehrer. 1982. Antonymy. Linguistics and Philosophy 5. pp. 483-501.
Murphy, Lynne. 2003. Semantic Relations and the Lexicon: Antonymy, Synonymy and Other
Paradigms. Cambridge: Cambridge University Press. pp. 169-203.
Partee, Barbara H. 2010. Privative Adjectives: Subsective Plus Coercion. Presuppositions and
Discourse: Essays Offered to Hans Kamp, eds. Rainer Bäuerle, Uwe Reyle, and Thomas
Ede Zimmermann. Bingly: Emerald Group Publishing Limited. pp. 273-285.
11